1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2010 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
22 #include <sys/mount.h>
32 #include "mount-setup.h"
33 #include "dev-setup.h"
41 #include "path-util.h"
48 typedef struct MountPoint {
57 /* The first three entries we might need before SELinux is up. The
58 * fourth (securityfs) is needed by IMA to load a custom policy. The
59 * other ones we can delay until SELinux and IMA are loaded. */
60 #define N_EARLY_MOUNT 4
62 static const MountPoint mount_table[] = {
63 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
64 { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
65 { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
66 { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
67 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
68 { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, false },
69 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
70 { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, false },
71 { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
74 /* These are API file systems that might be mounted by other software,
75 * we just list them here so that we know that we should ignore them */
77 static const char ignore_paths[] =
78 /* SELinux file systems */
81 /* Legacy cgroup mount points */
84 /* Legacy kernel file system */
86 /* Container bind mounts */
94 bool mount_point_is_api(const char *path) {
97 /* Checks if this mount point is considered "API", and hence
98 * should be ignored */
100 for (i = 0; i < ELEMENTSOF(mount_table); i ++)
101 if (path_equal(path, mount_table[i].where))
104 return path_startswith(path, "/sys/fs/cgroup/");
107 bool mount_point_ignore(const char *path) {
110 NULSTR_FOREACH(i, ignore_paths)
111 if (path_equal(path, i))
117 static int mount_one(const MountPoint *p, bool relabel) {
122 /* Relabel first, just in case */
124 label_fix(p->where, true, true);
126 if ((r = path_is_mount_point(p->where, true)) < 0)
132 /* The access mode here doesn't really matter too much, since
133 * the mounted file system will take precedence anyway. */
134 mkdir_p_label(p->where, 0755);
136 log_debug("Mounting %s to %s of type %s with options %s.",
147 log_full(p->fatal ? LOG_ERR : LOG_DEBUG, "Failed to mount %s: %s", p->where, strerror(errno));
148 return p->fatal ? -errno : 0;
151 /* Relabel again, since we now mounted something fresh here */
153 label_fix(p->where, false, false);
158 int mount_setup_early(void) {
162 assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
164 /* Do a minimal mount of /proc and friends to enable the most
165 * basic stuff, such as SELinux */
166 for (i = 0; i < N_EARLY_MOUNT; i ++) {
169 j = mount_one(mount_table + i, false);
177 int mount_cgroup_controllers(char ***join_controllers) {
183 /* Mount all available cgroup controllers that are built into the kernel. */
185 f = fopen("/proc/cgroups", "re");
187 log_error("Failed to enumerate cgroup controllers: %m");
191 controllers = set_new(string_hash_func, string_compare_func);
197 /* Ignore the header line */
198 (void) fgets(buf, sizeof(buf), f);
204 if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) {
209 log_error("Failed to parse /proc/cgroups.");
219 r = set_put(controllers, controller);
221 log_error("Failed to add controller to set.");
229 char *controller, *where, *options;
232 controller = set_steal_first(controllers);
236 if (join_controllers)
237 for (k = join_controllers; *k; k++)
238 if (strv_find(*k, controller))
244 for (i = *k, j = *k; *i; i++) {
246 if (!streq(*i, controller)) {
249 t = set_remove(controllers, *i);
262 options = strv_join(*k, ",");
270 options = controller;
274 where = strappend("/sys/fs/cgroup/", options);
286 p.flags = MS_NOSUID|MS_NOEXEC|MS_NODEV;
289 r = mount_one(&p, true);
298 if (r > 0 && k && *k) {
301 for (i = *k; *i; i++) {
304 t = strappend("/sys/fs/cgroup/", *i);
311 r = symlink(options, t);
314 if (r < 0 && errno != EEXIST) {
315 log_error("Failed to create symlink: %m");
329 set_free_free(controllers);
338 const struct stat *sb,
340 struct FTW *ftwbuf) {
342 /* No need to label /dev twice in a row... */
343 if (_unlikely_(ftwbuf->level == 0))
346 label_fix(fpath, false, false);
348 /* /run/initramfs is static data and big, no need to
349 * dynamically relabel its contents at boot... */
350 if (_unlikely_(ftwbuf->level == 1 &&
352 streq(fpath, "/run/initramfs")))
353 return FTW_SKIP_SUBTREE;
358 int mount_setup(bool loaded_policy) {
360 static const char relabel[] =
361 "/run/initramfs/root-fsck\0"
362 "/run/initramfs/shutdown\0";
368 for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
369 r = mount_one(mount_table + i, true);
375 /* Nodes in devtmpfs and /run need to be manually updated for
376 * the appropriate labels, after mounting. The other virtual
377 * API file systems like /sys and /proc do not need that, they
378 * use the same label for all their files. */
380 usec_t before_relabel, after_relabel;
381 char timespan[FORMAT_TIMESPAN_MAX];
383 before_relabel = now(CLOCK_MONOTONIC);
385 nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
386 nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL);
388 /* Explicitly relabel these */
389 NULSTR_FOREACH(j, relabel)
390 label_fix(j, true, false);
392 after_relabel = now(CLOCK_MONOTONIC);
394 log_info("Relabelled /dev and /run in %s.",
395 format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel));
398 /* Create a few default symlinks, which are normally created
399 * by udevd, but some scripts might need them before we start
403 /* Mark the root directory as shared in regards to mount
404 * propagation. The kernel defaults to "private", but we think
405 * it makes more sense to have a default of "shared" so that
406 * nspawn and the container tools work out of the box. If
407 * specific setups need other settings they can reset the
408 * propagation mode to private if needed. */
409 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
410 log_warning("Failed to set up the root directory for shared mount propagation: %m");
412 /* Create a few directories we always want around */
413 mkdir_label("/run/systemd", 0755);
414 mkdir_label("/run/systemd/system", 0755);